Currently, the TESMTH function (optionally) can compute the optimal values for the smoothing parameters. The user has the option to pass the initial values for the optimizer or leave it blank and the function assumes a constant defaults (e.g. 0.333).
In a paper by Chatfield and Yar (Journal of the Royal Statistical Society. Series D (The Statistician), Vol. 37, No. 2) , they authors recommend (1) using default values that are data-driven, and (2) normalizing of the seasonal indices.